Search

Due to the diversity of power consumption patterns, the false positive rate (FPR) of data-driven electricity theft detection (ETD) methods is too high to meet practical needs, which severely restricts the engineering application of data-based methods. To reduce FPR of ETD methods based on advanced metering infrastructure (AMI), a deep neural network with low FPR (LFPR-DNN) is proposed in this paper. First, a deep model is constructed based on one-dimensional convolution and residual network, which can automatically extract features from consumption data. Then, a two-stage training scheme is used to train the network. In the first stage, the conventional gradient descent algorithm is used to update the network weights. To minimize the impact of data imbalance on detection performance, focal loss is used. Besides, grid search is used to optimize the hyper-parameters of the model. In the second stage, with FPR as the optimization objective, the particle swarm optimization (PSO) algorithm is used to train the network. Finally, the proposed LFPR-DNN is verified by using the open Irish data set. Compared to other state-of-the-art classifiers, LFPR-DNN has the lowest FPR with 0.29% and the highest AUC with 99.42%. The FPR is reduced by an order of magnitude, which verifies the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Nowadays, electricity theft has been a major problem worldwide. Although many single-classification algorithms or an ensemble of single learners (i.e., homogeneous ensemble learning) have proven able to automatically identify suspicious customers in recent years, after the accuracy of these methods reaches a certain level, it still cannot be improved even if it continues to be optimized. To break through this bottleneck, a heterogeneous ensemble learning method with stacking integrated structure of different strong individual learners for detection of electricity theft is presented in this paper. Firstly, we use the grey relation analysis (GRA) method to select the heterogeneous strong classifier combination of LG + LSTM + KNN as the base model layer of stacking structure based on the principle of the highest comprehensive evaluation index value. Secondly, the support vector machine (SVM) model with relatively good results of the stacking overall structure experiment is selected as the model of the meta-model layer. In this way, a heterogeneous integrated learning model for electricity theft detection of the stacking structure is constructed. Finally, the experiments of this model are conducted on electricity consumption data from State Grid Corporation of China, and the results show that the detection performance of the proposed method is better than that of the existing state-of-the-art detection method (where the area under receiver operating characteristic curve (AUC) value is 0.98675). [ABSTRACT FROM AUTHOR]

• The multi-task learning is used to simultaneously identify electricity theft users and detect their electricity theft time period. • The proposed model named MDRN is designed by using residual network, and the hyper-parameters of the model are selected by grid search. • A multi-task loss with task uncertainty is proposed to automatically balance the weights of two tasks during the training process, which improves the comprehensive performance of the multi-task model. • The effectiveness of the proposed method is verified using the Irish data set. To demonstrate the excellent performance of the method, we compare the proposed method with popular deep learning (DL) algorithms. Electricity theft is an economic problem faced around the power companies. After the suspicious electricity theft users are recognized, the potentially stolen electricity (PSE) and the loss caused by the electricity theft needs to be accurately evaluated. In the actual inspection work, the key to evaluating PSE is electricity theft time period detection. However, most of the current electricity theft detection algorithms can only recognize electricity theft, which cannot detect the specific time period of electricity theft simultaneously. Therefore, this paper proposes a deep model named multi-task deep residual network (MDRN), which can simultaneously recognize electricity theft users and detect their electricity theft time period. The MDRN is constructed based on one-dimensional convolutional and residual network, which can effectively extract characteristics from power consumption data. To automatically balance the multiple tasks in training, a joint multi-task loss with task uncertainty is proposed. The experimental results based on the Irish dataset show that the proposed multi-task model obtains the highest Accuracy with 93.17% in electricity theft recognition and the highest intersection-over-union (IOU) with 76.58% in time period detection. It should be noted that the proposed method can be directly used to calculate the PSE to maximize economic return of electricity theft inspection. [ABSTRACT FROM AUTHOR]

With the establishment of advanced metering infrastructure (AMI), more and more deep learning (DL) methods are being used for electricity theft detection. However, there are still problems such as the construction of a DL model that better fits the theft detection task, sample imbalance in the data set and overfitting of the model which limit the potential of DL models. Therefore, a novel end-to-end solution based on DL to solve these problems for electricity theft detection is proposed in this paper. First, we construct a model based on Transformer Neural Network (TNN), which extracts global features of consumption data by calculating the self-attention between each segment obtained by dividing the whole load sequence, and calculates the relative relationship between features for customer classification. Then, we refine the model to further improve its performance. Conv-attentional module is used to embed the input data and capture the local features in each segment. And we minimize the effect of sample imbalance by choosing a suitable loss function, and address the problem of model overfitting by adding normalization layers, dropout regularization and L2 regularization. In addition, grid search is used to determine the optimal values of the model hyper-parameters. Finally, the performance of the proposed model is verified by experiments using the Irish data set. The results show that our method is able to extract features more efficiently and thus has a higher true positive rate (TPR) with a lower false positive rate (FPR) than other state-of-the-art detectors, and it has strong robustness. • Construct an improved Transformer network-based model to detect electricity theft. • Use conv-attentional network to embed electricity consumption data. • Eliminate the effect of class imbalance by selecting a suitable loss function. • Solve the problem of model overfitting by adding various network structures. • Conduct sufficient experiments to verify the performance of the proposed model. [ABSTRACT FROM AUTHOR]

• There is a serious sample imbalance in the data, which easily leads to a severe bias in the classification model, which in turn affects the classification accuracy. In order to solve the problem of sample imbalance, we proposed a method of using focal loss for model training to enhance the contribution of a small number of samples to model optimization. • Electricity consumption data is correlated in the time dimension. It is necessary to capture a larger receptive field to obtain global information and improve the model's accuracy. Keeping this concern in mind, we used dilated convolution to enlarge the receptive field without loss of information so that each convolution output contains a larger range of information, thereby capturing a larger receptive field and obtaining global information of the input data. • How to effectively integrate the feature expressions obtained by wide CNN and deep CNN is a key factor to improve the performance of the model. To address this problem, we adopt the channel-dimensional adaptive attention module to adaptively fuse the feature expressions obtained by wide CNN and deep CNN to improve the accuracy of model training. For the increasingly serious phenomenon of electricity theft, many researchers are trying to detect it. Traditional detection methods rely on physical inspection, which has low detection efficiency and high cost. However, the rapid application of advanced metering infrastructure (AMI) makes it possible to detect electricity theft via smart meters. In this work, we propose a hybrid improved wide and deep convolutional neural networks (CNN) method to address electricity theft detection on an imbalanced realistic dataset that presents a daily electricity consumption provided by State Grid Corporation of China. Our key contribution is the introduction of a channel-dimensional adaptive attention module concatenated with dilated convolutions. Moreover, we propose focal loss to solve the data imbalance problem. Experimental results demonstrate that compared to other state-of-the-art deep learning classifiers, our improved method has better performance (with 97.08% of MAP@100 and 83.61% of AUC), which verifies the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]